A-16394 RESEARCH AND DEVELOPMENT AT THE CENTER FOR SEISNIC 1/1STUDIES(U) SCIENCE APPLICATIONS INC NCLEAN VAC F ROMNEY ET AL. 14 NAR 86 SRI-96168

UNCLASSIFIED MDA9S3-84-C-062S F/G 8/11 M

Em hhhhhhh- miuEmIIIIIIIIIIIIIi flflflfl..f

V ||||aa

111la 1.0 M 32

1 1.5 14 "

MIr-YSIII TMTS HR

L. *3

I , Technical Report C86-02March 1986

is

QUARTERLY TECHNICAL REPORT

FOR 1 APRIL 1985- 30 JUNE 1985

Center Staff DTIC

-S APR~ 10

i:& public leLease and sale; its

dicustib,.±tion is unlimited.

SPONSORED BY:

DEFENSE ADVANCED RESEARCH PROJECTS AGENCY

I L1300 N. 7th Street, Suite 1450'

Afrgo Vrn 02209-71Teepone: (703) 2 7900j 4 10 005 :.

Technical Report C86-02[m "March 1986

QUARTERLY TECHNICAL REPORT

FOR 1 APRIL 1985 - 30 JUNE 1985

p: Center Staff

.

The views and conclusions contained in this document are those of the authors andshould not be interpreted as representing the offical policies, either expressed orimplied, of the Defense Advanced Research Projects Agency or the U.S. Government.

Sponsored by:DEFENSE ADVANCED RESEARCH PROJECTS AGENCYMonitored by: Science Applications International Corporation" Defense Supply Service - Washington 1735 Jefferson Davis Highway. Suite 907 ,.

Under Contract No. MDA 903-84-C-0020 Arlington, VA 22202

SDefens Sup Serv -Washigto n 173-5er-aH'.St0.

L Uncl assi fiedSECURITY CLASSIFICATION OPFTIS PAGE (*Wona Date alo 00e.d

REPORT DOCUMENTATION PAGE READ__ INSTRUCTIONS _

I. __ ___REPORT_ __ __ ___NUMBER_ __ __ _ BEFOREC0O6PLETINGFORM

C R N U1E . G O V T A C C E SSIO N N O . 3. R E C IP IE N T'S C A T A L O G N U MBl E Rl

Ot 4 TILE And ub~le)S. TYPE or REPORT & PERIOD COVEREO

Research and Development at the Center Quarterly Technical Report

for Seismic Studies 1 April 1985- 39 June 1985~ 6. PERFORMING ORG. REPORT NUMBER

AUTHR~s)SAIC -86/10681- ~ .AUTOR~,I . CONTRACT OR G~ T NumBERI.I

Carl Rominey Alan Ryall MAO-4 02 NGerald Frazier Michael TiberiaHans Israelsson ______________

3. PIERFORMING ORGANIZATION NAME ANO ADDRESS 10. PROGRAM ELEMENT. PROJECT. TASK ofAREA & WORK UNIT NUMBERS

Science Applications International Corp.~ 1710 Goodridge Drive

McI an Virginia 22102 _______________

CONTROLLING OFFICE NAMIE AND ADDRESS 12. REPORT DATE

Defense Advanced Research Projects Agency 14 March 1986 21400u Wilson Boulevard 13. NUMBER Of PAGES

Arlington, Virginia 22209 3114. MONITORING AGENCY NAME a AOORESS(i1 diffeeant 110dm Controlingj Office) IF. SECURITY CLASS. (of this .ePort)p Defense Supply Service - Washington

Room 1D245, The Pentagon Uncl assi fiedWashington, D.C. 20310 I~.OCASIFICATION/OOWNGRON

SCHEDOU LE

16. DISTRIBUTION STATEMENT (of this e to) ..

Unlimited

17. DISTRIBUTION STATEMENT (of theabatract aftr~iee n Block 20, Il dlf.,aflt fromn Report)

Unlimited

* . III. SUPPLEMENTARY NOTIES -

I9. KEY WORDS (Continue on fewer** aide if nocoeeery ad fdontlfy by block num ber)

Seismology.-- Magnitude 3ias.Data Center,Seismic Data Exchange,Monitoring Nuclear Tests,

A. 20. AU@$TRACT (Continue an everse cede It necessary e~ Idenltf by block number)

",This report covers results from an international seismic data exchange testsponsored by the Group of Scientific Experts, U.N. Conference on Disarmament.It also reports on computer program development at the Center for SeismicStudies, on mb bias as deduced from Soviet seismograms, and on noise levelsand detection thresholds of seismic stations that particapated in the dataexchange experiment. ±:

DD I jAH 7 1473 EDITION oF I NOV SS IS OBSOLETE Datessi ieSECURITY CLASSIFICATION OF TmiS PAGE (*%on en eLed

1. FOREWORD

2. GSE TECHNICAL TEST

2.1 Summary of Activities

2.2 Routine Data Processing Software Developed for GSETT

2.3 Formation of the Combined WashlDC/ShocIDC GSETT Database.. 4

92.4 Upgrades to the International Seismic code

S 2.5 Further Analysis of Level I Data Transmissions Received over WEO/GTS

2.6 GSETT Network Detection Thresholds Inferred from Earthquake

Recurrence Curves

3. SUN DEVELOPMENT

% " .. 3.1 New Data Retrieval Software on the SUN Microsystem

3.2 Center Standard Software Now Available on the SUN

4. RESEARCH

4.1 Preliminary Study of mb Bias at Selected Soviet Seismic Stations

4.2 Noise Values and detection Thresholds of GSETT Seismologial Stations

IiF

* V . . . . -.- . . •. .

°,.,.,,.,,,%,,,,2 . . .. ' :. ' ,' .. .'. ,' . ,' W,' r ,. . ,2 '.' ,' ¢ . '% ', -,,. ,, -, .,,., -.' '..-- .V-. *.,.,-.. ,..'. ....

1. FOREWORD

Commencing about December 1983, and continuing through the period covered bypreparing for, and conducting the major part of the United States' technical

activities associated with an International Seismic Data Exchange experiment. Asdescribed in previous reports, virtually all work conducted under this project bythe scientific staff of the Center was devoted to this project, known as theGroup of Scientific Expert's Technical Test, or GSETT. The main period ofoperational activities for the GSETT was 15 October 1984 through 15 January 1985,and the major descriptive documentation of the Center's activities was completed

IN by March 1985. The reader is referred to the proceding quarterly report for more

N detail on Center-specific activities, and to the forthcoming report of the Groupof Scientific Experts, UN Conference on Disarmament, for information on the testas a whole.

} ,: During this quarter work continued on documentation of the GSETT and onperforming certain routine assessments, as will be described in Section 2 of thisreport. However, with the reduction in demands by the GSETT it was possible todirect a portion of the technical staff's efforts towards other objectives, asdescribed in Sections 3 and 4 of this report. Section 3 outlines steps taken byMike Tiberio to make the SUN 2 computers at the Center more useful for conductingseismic analysis. Section 4 gives abstracts of research conducted by members ofthe staff in two areas. Alan Ryall's short report on mb bias at Soviet seismicstations documents work initiated at the University of Nevada and completed at

, P the Center as a visiting scientist and later as a member of the staff.Hans Israelsson's short report on noise levels and detection thresholds at GSETTstations represents the initial steps of a larger research project directedtoward exploiting the excellent GSETT database for research on a variety ofproblems of interest to nuclear test monitoring. Both of these topics are thesubject of separately published detailed reports, C86-03 and C86-04.

Aeos'onF or

NTIS CRA&IOTIC TABUnannounced 0Justificatin.

. .Av a ila b ility Co d s a'T

Avail aidl Ior-PN

"F. . . . °...

- . . . . .. . . . . o . . . . . . .. . . - . . * , .- . - .- . .i -. - . . o *. .

: ' .. ' -. ',€ -. ,' ',, '. .,: r. ,o',,.'.' .. ,' .- '.. '_,.,' '.- . '.' - -2,' ,.' " '.'. /,-' L01,st ... r_..t ., .e,.. _i", ."

Igi

I2. GSE TECHNICAL TEST (GSETT)

3 2.1 Summary of Activities

The GSETT was completed in a formal sense during the previous quarter (on 15January 1985), and rather complete reports describing the experiment and itsmajor results were prepared for use by the U.S. delegation at the March meetingof the Group of Scientific Experts (GSE). However, a number of follow-upactivities continued throughout this quarter that can be considered either as *':,part of the experiment itself, or as part of the documentation of theexperiment. Some activities of this nature were initiated in response to needsof the GSE in preparing its formal report to the Conference on Disarmament, andprocess of conducting more thorough evaluations of the experiment. Highlights of

these activities included:

- A tape was received from the Experimental International~~Data Center in Stockholm, Sweden, containing the completeset of Level I data messages received during the GSETT.

Programs were written to merge this new data with thatreceived at the Center (after purging duplicates).The new consolidated data set was then parsed and Installed:h i in an Ingres database.

- Using this combined database, analyses were conductedof the reported noise levels from the participating GSETTstations. Individual station "spectra" (average noiseamplitudes at each period) were prepared. These resultswere then supplied to Dr. Robert North, Energy, Mines

P 1and Resources Canada, for inclusion in the GSE report.

- Logs of data messages received at the Moscow ExperimentalF" ,"International Data Center were obtained. The information containedp. .therein, together with the tape from Stockholm,

provided a complete list of all messages received at thecombined EIDCs.

- Analyses were conducted of the combined Level I databaseto produce statistics on transmission effectivenessof the Global Telecommunications System of the WorldMeteorological Organization. These statistics showedthe numbers of messages received on the firsttransmission, as well as those received after retransmission, ,from each participating country. Programs were also writtento determine the total amount of seismological datareceived, based on "begin" and "end" statementscontained in the messages.

25,

-.5 .'...%S V '

- Studies were initiated to evaluate the adequacy ofthe automatic event detection (the "ODP" and "PDPO)programs employed by the Center during the GSETT.As part of these studies, programs were written toNwindow" waveform segments containing P-Waves,to scale them arbitrarily in amplitude, and to blendsignal and noise samples.

-Preliminary analyses of the detection threshold ofindividual stations and of the total GSETT networkwere conducted. .-

Im "Details on several of these and other activities are contained in the following

portion of Section 2 of this report.

1~~ V

J

,%, *e-i

r.

-.*'%

V. :"--"

3 ",

B

2.2 Routine Data Processing Software Developed for GSETT

During the GSETT experiment the Center was run for the first time as a fullyoperational seismic data processing center. While most of the basic modulesnecessary to perform the required technical analysis had been developed over theyears, many scripts had to be written for managing and preprocessing the data forproper input to the analysis programs and to insure a smooth operation. Inaddition to these scripts, some new software had to be written to handle thoseaspects that were unique to the test (i.e., Level I parameters). This reportoutlines those scripts and programs that were written solely for the test, and

explains how they fit into the data flow and how they work.

The following table identifies these scripts and programs and provides anindication as to where, why and how each was used.

NAME ?DC SCRIPT MACHINE FUNCTION

V SUN3 ndc yes vax capture wf's, run

11. build tapesScutarr ndc yes vax eliminate RSNT and

- nRSCP from flowresu. ndc yes vax build archive

tape request for sunarr2sunreq ndc yes vax generate sp/mp

request from arrivalsy- ndc yed vax generate continous

lp requestsunarc ndc no vax write archive tape

in SUN integers

12 ndc no vax generate Level Iparameters

memup ndc yes sun make directoriesI. and get arrivals

dmparc ndc yes sun check disc arrivals,run sundmp

GPM ndc yes sun drives gpm programgpm ndc no sun graphical

parameter measurementpp ndc no sun post parameter processingsend ndc no sun generate wmo messages

idc yes vax check for unparsedmsgs, run parser

-** CHK idc yes vax check qualityof parser output

MAD idc yes vax make a script to load.* "'-parsed data'. OTH idc yes vax find and parse telex

messagesMAA idc yes vax make a script to load

i telex data

4TT :'

To function as a "National Data Center" it was necessary to tap into ourroutine archives to obtain data. SUN3 is a very large script that performs manyfunctions. First it logs the standard continous data archive tapes.This allows the data to be segmented later. Once the detector and post detectorprocessing has been completed, the data can be segmented. This requires that thesegments be obtained from the continous archive tapes. Data from two of the fiveseismic stations received on-line at the Center, RSCP and RSNT, were not analyzedduring the test and the cutarr script removed them from the "dearchive" arrivalfile. The reqsun script bu-i-7ds this dearchive request. Re sun calls arr2sun:eqwhich turns thedetector calls into requests for short peiodT sz,sn,se) and mid-- "" period (mz,mn,me) data. The long-period data is not segmented, and 1preq builds "

a request for the continuous lp data. These segments are then written to anarchive tape using a special version of the archiver (sunarc) that writes SUNintegers on the tape rather than VAX integers. This saves the SUN from having todo it, and reduces the total amount of time, overall, for the data transfer.Lastly, before the segments are removed from the VAX, the Level I parameterprocessor (lip) is run. This program supplements the standard arrival file withrecords indicating estimates of MIX, M2X, NSZ etc. After this step the data is

* ,removed from the VAX and the data tapes carried to the SUN. The analyst, beforereading the tapes, would run the script se p in order to retrieve, over the netfrom the VAX, the necessary arrival files. Setup would also build the properdirectory tree structure so that when the tapes are read, directories will exist -j for the data. Once the tapes were mounted on the SUN tape drive, the analyst ranthe script d . _Dmn , before reading the tape, would check and make surethat sufficient room existed on the disc. Once the disc space checked out,dmparc would run the program sundmp, which would actually build the waveform

, database on the SUN. Once the database is built, GPM4 can be run. GPM the scriptruns gpm the program, taking care of some bookeepiig-along the way.•Afteranalys-Is, PH, the post parameter processor, was run, insuring that the work

. q. performed conformed to normal seismic rules and also that it conformed to rules --set out by the GSE concerning their special parameters. Once checked out, thescript send, which calls the program wino, generated messages W140/GTS format fromthe Centerdatabase files.

L To function as an "International Datacenter", numerous large parameter fileshad to be manipulated. What follows is a detailed description of the scriptswritten to handle these files. PRS is a script that would allow the user toparse incoming W* O seismic messages, either interactively or in batch mode.Normally the user would run the files through in batch mode, which would allowall the error free files to be parsed and would leave all the files containingerrors unparsed. The user would then parse the erroneous files interactively,editing out the errors so that the files would parse. PRS would first search for -messages with invalid or missing arrival files. This list of undone messages wasthen processed. After all the messages were parsed, the script CHK would check ,.

. the arrival file produced by the parser for common errors. Non-un--Tque arrivalidentification numbers or non standard record lenghts are sufficient to corruptthe database and thus were not allowed to pass.

5-J %.-.

Any messages producing such files were corrected and reparsed. Once all thefiles checked out it was necessary to load the files into the database. Thescript MAD produced a script to do that. Special scripts to handle telexmessagesand those messages whose headers were so corrupted that they could notbe classified, were needed. OTH and HAA perform tasks similar to PRS and MAD.

M. A. Tiberio

. .,.. ,

6

-, . *.. -. C- . .-

2.3 Formation of the Combined WashIDC/StocIDC GSETT Database

A new database called GSETT was created to store seismic parameters reportedduring the experiment. The database as it presently stands consists of a numberof tables that contain the parsed equivalents of the WMO messages received during

* the experiment. It was necessary to use four tables to hold the data containedin the GSETT messages. They are staog• arrival, feature and extra. Initiallythe GSETT database contained only the datat was received by-the WashingtonExperimental International Data Center (WashlDC). Also, only those messages thatwere deemed to be non-duplicates were loaded into the database. Messagesreceived prior to a cutoff time of 0600 UTC on the date that a Preliminary

., Epicenter List (PEL) was to be calculated were parsed once daily and loaded intothe database. Thus at the end of the experiment, the GSETT database containedall the WashIDC contributions. It was noticed that when message coverage was

* compared between the WashIDC and StocIDC, there were a number of messages- received by only one data center. The following table shows how many messages

were unique to each IDC and how many were jointly received. These numbersreflect total numbers received, not unique messages received.

# messages total # duplicated # uniqueWASHIDC 3023 3019 4STOCIDC 3330 3152 178

One can see that Stockholm received 174 more messages than did Washington. In .7]order to make a more complete GSETT database, the Stockholm dataset was obtainedand distributed into a directory tree that paralleled the one used during the

experiment.

Once the missing messages were identified, the duplicates within this listwere removed, resulting in 98 unique new messages being parsed and loaded intothe GSETT database. Most of the messages received by Stockholm but not byWashington were from SEAG (Argentina), SEAU (Australia), SECN (Canada), SEIN(India) and SEZB (Zambia).

M. A. Tiberio

r J

7 r

..................-. ...... .%•

2.4 Upgrades to the International Seismic Code

An important aspect of the GSETT was the design and implementation of manynew procedures. Conventional procedures for measuring signal parameters were

, ~* refined and new ones were added. Also, new message conventions were needed forreporting these data (Conference Room Paper 134/Rev .1, Procedures for the GSE ___

Technical Test, 1984). Many of the specially designed procedures used during thetest have application to conventional seismological practice. This sectionreports one such example, namely improved message conventions for reportingseismic data.

Following the GSETT, Drs Gerald Frazier from the Center for Seismic Studiesand Peter McGregor, the test coordinator from Australia, met with the WorkingGroup on Telegraphic Formats, Subcommission on Data Exchange of the IASPEI

Commission on Seismological Practice to upgrade standard message conventions,i.e., the International Sesmic Code (ISC). The 1985 version of the ISC resultedfrom this meeting, subsequent correspondence and the considerable efforts ofJerry Dunphy and his colleagues at the U.S.G.S. National Earthquake InformationCenter who prepared the actual code documentation. This new code accommodates

essentially all of the special GSE parameters and includes the benefit of furtherrefinements identified during the test.

The special procedures used for encoding GSETT messages were derived from thethen current convention, ISC version 1979. The GSE message conventions weredeveloped in an attempt to specify unambiguous message formats in a way that wasnot easily misunderstood, and to make provisions for all relevant data. Tominimize conflicts with the current conventions, several of the GSE extensionswere implemented as comments, isolated from the formatted data by double sets ofparentheses. These unobtrusive extensions to the message convention includedsuch relevant data as: message identification (i.e., GSETT), message originator,identification of special message types (i.e., bulletins), time interval coveredby each station report, data outages during this interval, and the analyst'sinterpretation of recorded events. With the exception of the analyst'sinterpretation of recorded events, the 1985 version of the ISC now accommodatesthe above information in the formatted portion of the message rather than incomments as was done in the GSETT. Concern was expressed that the particular setof GSE key words that were used to enter the analyst's interpretations could leadto omissions of relevant measurements, and this procedure was not adopted.

In preparation for the test, new phase identifiers were defined for encodingnon-standard data, and provisions were made for identifying the recording channel

4from which the data were extracted. Special effort was required to providemeaningful, unambiguous identifier names. Among the alternatives that wereconsidered was an attempt to append letters to the conventional phase identifiersto explicitly specify the recording channel, i.e., the instrument response band(S,M,L) and the directional component (Z,N,E). However, we found that, whenthese two-letter combinations (SZ,SNSE,I4Z,14E,LZ,LN,LE) are appended as suffixesto the common phase identifiers (P or S), over half of the resulting three-lettercombinations coincide with one of the many station abbreviations. Thisparticular scheme was abandoned in favor of an alternate procedure that avoidedthese ambiguities. Special characters have now been defined in the new ISC tocompletely specify the channel, both the frequency band (i.e., instrument class:SP,LP,MP,BP,UP) and the directional component (Z,N,E).

8

ZZ.1

Further ambiguities with station abbreviations were noted. Half of thethree-letter abbreviations used to denote the month-of-year coincide with astation abbreviation, and message context is required to distinguish the two•.(The 1985 version of the ISC has mostly alleviated such ambiguities by requiring:

"A colon (:) must be prepended to a station or netabbreviation whenever the abbreviation is identicalto a phase code or symbolic identifier used in the

S'.International Seismic Code."

This action will simplify the computer task of decoding messages and enhance -thereliability of data exchange. Added reliability could have been achieved byrequiring that a colon be prepended to all station names. However, such arequirement was not made in the new ISC so as to maintain compatibility withmessages prepared according to the earlier 1979 code, which has no provisionalcharacters for identifying stations apart from their unique abbreviations.

The desirability of special characters (e.g., : or /) and message delimiterswas recognized during the planning stage of the test, and some provisions weremade. Again, to minimize conflicts with the current conventions, no specialcharacters were used, but spaces, carriage returns and double carriage returnsplus station abbreviation and date were required to separate parameters, phases,

"- and events, respectively.

* Execution of the test enabled us to evaluate the special message conventionsand to identify further improvements. An average of about 50 messages weredecoded per day using an interactive computer parser developed at the Center in %-.70

* preparation for the test (see the preceding quarterly technical report). The-" parser was designed to perform a sequence of tests to verify the identification

of parameters in the message. When problems were encountered, the parserdisplayed the message contents in the near vicinity of the problem, identifiedthe discrepancy, and placed the analyst's keyboard in an edit mode with the

. -. screen cursor positioned on the apparent error. Otherwise, the parser proceededto decode messages and write the data in database-formatted files, operating in afully automatic mode.

To resolve non-unique constructs the parser was equipped with numerouscontext flags to portray recognizable information before and after the ambiguous

- 'data. The required spaces, carriage returns and station delimiters proved quiteuseful for identifying local context and to aid in resolving potentialambiguities. Overall, the messages were well formatted, and the participantsshould be commended for the care that was taken to implement the newprocedures. Nevertheless, nearly half of the messages contained an irregularitythat required analyst intervention to fix. The large majority of these computer-identified problems were resolved either by accepting the algorithmic assumptionsor by implementing obvious fixes, e.g., dropped or substituted characters (e.g.,an"I" used for a one or an "0" used for a zero). About two hours of an analyst'stime was generally required each day to decode the messages depending on thenumber of irregularities that were encountered and the computer load from other

L jobs.

' ,... ,

9 **:

• 4

Pathological cases were encountered that were not easily resolved, and addedlevels of sophistication were implemented during the test to handle some of the

-r peculiar constructs. Spaces, inserted or omitted at unusual places within themessage, posed a particular challenge. To automatically decode such data iscumbersome and unreliable, and correcting this data involves experienced humanjudgement. It became apparent that much of this awkward dilemma could have beenavoided through the use of special characters to delimit sub-message constructs.

The 1985 ISC has taken important steps to improve the message formats used° "for exchanging data. In addition to the extensions already noted, the new code

specifies spaces and carriage returns to delimit several types of data, and acomma (,) is now used to separate first motion parameters measured on short

-; -- period and long period channels. Special characters have been designated to kdelimit reported events (BEGEV, ENDEV and /). Several new phases have beendefined, akin to the GSE parameters, including maximum signal amplitude and pre-signal measurements of ground noise.

Overall it appears that the GSETT has provided valuable information forupgrading the standard ISC message conventions. Although many of the new

41 ,; features are optional, their universal adoption would greatly simplify the taskS.of message decoding and provide improved reliability for exchanging seismic data. -,

G. Frazier

6k.J

:2 "2

.7 .,

10 r. .'."..' -

*

%2.5 Further Analysis of Level I Data Transmissions Received over WNO/GTS

The preceding quarterly technical report contained rather complete reports ofGSETT messages received over the Global Telecommunications System of the WorldMeteorological Organization (WMO/GTS). While useful for generating statistics ontransmission efficiencies, these reports did not give a clear picture of acrucial aspect of the test: how much seismic data was actually available at theExperimental International Data Centers (EIDC) for use in detecting seismicevents?

To investigate this question, programs were written to determine, for eachcontributing station, the time intervals for which data was received by one ormore of the EIDCs. Station outages of less than 6 hours were not counted as themissing for purposes of this study. The results were then used to determine thepercentage of time during the test for which data were received from eachstation, as well as the percentage of data received for each day during the test.

The attached table gives the percentage of time for which data reported froma participating station was received over the WMO/GTS at any of the threeEIDCs. For over one half of the stations participating in the GSETT, reportswere received which covered at least 95% of the time of the test. As expected,

- there is a strong correlation between those stations having a large fraction oftheir station data missing and the less efficient circuits of the WMO/GTS as

- revealed in the study included in the preceding quarterly technical report. Theeffective transmission rates for South America, Africa and New Zealand neverexceeded 20%, even accounting for retransmission which had an insignificanteffect on these circuits. Loss of this data is of particular concern in view ofthe sparsity of seismic stations in these regions. The combined effect was a

pm relatively poor signal detection capability for these regions during the GSETT.

Figure A is a chart of the overall availability of the combined set of datafrom the GSETT network received at one or more of the EIDCs. The daily average

.availability of data from the stations in the whole GSETT network was 72% duringthe Test. The data on which these statistics are based was compiled at theconclusion of the GSETT; thus the percent availability indicated represents a

. rm a x im um a f t e r r e t r a n sm i s s io n a n d i s n o t in d ic a t iv e o f t h e am o u n t o f d a taavailable for calculation of the Preliminary and Final Event Bulletins, which mayhave been considerably less. A surprising result is that the overall

- availability of data at the EIDCs did not improve during the GSETT. This lack ofim provement is ind icative of chronic problem s w ith certa in W14O /GTS c ircuitsduring the test due to either poor connections with national facilities or poorquality communication circuits.4.,*

- ..4.:

T;

I'

r •

. o".,

LEVEL I DATA FROM STATIONS(VIA WMO/GTS COMMUNICATIONS)

_ __,_ October 27 - December 14, 1984Country/Station No. of Days Percent Country/Station No. of Days Percent

' _ _ _ _ of Level I Data Coveragell of Level I Data Coverare.ARGENTINA 'MJIA

BAA 5.0 10 GBA 47.2 96LPA 10.5 21 INDONESIA

AUSTRALIA AAI 26.5 54ASPA 49.0 100 JAY 7.7 16CTAO 48.2 98 KSI 6.0 12 ,MAW 48.2 98 KUG 14.0 29 y

NWAO 47.7 97 PSI 17.7 36 -K. .AUSTRIA rRT 15.5 32

" ' KBA 49.0 100 IRELANDBELGIUM DKM 47.2 96

DOU 47.5 97 ITALYUCC 46.2 94 MNS 46.5 95

BRAZIL RMP 46.5 95BDF 17.5 36 JAPAN

BULGARIA MAT 48.2 98CANAD 42.7 67 NEHERLANDS

CND BN 49.0 100GAC 49.0 100 ENN 49.0 100MBC 49.0 100 NEW ZEALAND '

YKA 49.0 100 RAR 6.5 13COLOMBIA BA 6.2 13

BOG 0.7 1 WEL 2.5 5CZECHOSLOVAKIA ORWAY "

KHC 49.0 100 NAO/NB2 49.0 100PRU 49.0 100 PERU

DENMARK A 28.2 58COP 41.7 85 ROMANIADAG 41.0 84 MLR 19.7 39GDH 28.5 58 SWEDEN

. EGYPT HFS 49.0 100 •

HLW 4.2 9 APO 49.0 100FINLAND SLL 49.0 100

NUR 49.0 100 TBY 49.0 100SUF 49.0 100 U. S. S. R.

FRANCE OBN 49.0 100LOR 44.0 90 U. K.PMO 11.7 24 EKA 48.0 98

GERMAN DEM. REP. U. S. A.MOX 47.5 97 LAC 49.0 100

GERMANY, FED. REP. LTX 48.0 98GRF 49.0 100 FBAS/FBAL 49.0 100

*1 HUNGARY RSNY 49.0 100BUD 45.2 92 RSON 49.0 100

JOS 45.2 92 RSSD 49.0 100PSZ 45.2 92 ZAMBIA

._- LSZ 57.2 76

Outages within message periods are not included.

Total Level I data from each station based on data set combined from all EIDC'sat conclusion of GSETT.

12

r

,'? ".,.-

There was an appreciable difference in the average transmission effectivenessof the WMO/GTS circuits in the Northern Hemisphere as compared to those in theSouthern Hemisphere. This is directly reflected in the availability of data atthe EIDCs from stations located in these areas. Figures B and C show the overall j" *

data availability, after accounting for retransmission, at one or more of theEIDCs from stations in the Northern Hemisphere and the Southern Hemisphererespectively. It is seen that, on the average, 89% of the data from stations inthe Northern Hemisphere were received by at least one EDIC. For stations in theSouthern Hemisphere, the average data availability was only 44%. By contrast,those national facilities from which more than 95% of the messages were received

* -' were almost exclusively in the Northern Hemisphere. Australia was the onlyexception to this. The efficiency of the W4MO/GTS and the high density ofstations in North America and Europe resulted in a relatively good signaldetection threshold for these regions.

L. HuszarC. Romney

.0

13

A. .. . . . . . .

~..............

.. . . . . .. . . . . . . . . . . . . . . . . . . . . . .

.9

ir.1

IIs

" ' l~~igumi A Ovrll| Network Data Availability" of Comlmbit,4 EM'I), Vertical,' -

q • olummos indicate the percentage of participating statins whoss

.. ... .. ..

cull 1. UIX I.T Plin a. c ,

1-0

?a,

IF;

~~14

CM11 Ift* LIMI Pl MiCteW, W IK*S

ii

.. .... ....

ftttm (Inluin Bool

I15i . -.- '

as

. :. ..-.,.

' ][r~~Ii. C Dila Awaikbilily o( Combised EIXDc. f,.. hoghrs, Uemiplbe "-

II..

;515

[,',,

-, -

2.6 GSETT Network Detection Threshold Inferred from Earthquake Recurrence Curves

Final Epicenter Bulletins from the Experimental International Data Center(EIDC) in Washington, Stockholm and Moscow were calculated and transmitted to allparticipants during the GSE Technical Test (GSETT). According to the GSETTinstructions, all EIDCs should have used a common program for associating theevents reported by the participating stations and for calculating the formalparameters of located events. Given the same data reports, the calculatedepicenter lists should have been identical. Some of the associated reports would mT'have resulted in spurious "events" because of imperfections in the AutomaticAssociation (AA) program, but with coordination among the three EIDCs, presumablythese could have been weeded out, and identical final epicenter lists would havebeen produced.

In practice, there were differences in the analysis programs among all EIDCs(although programs used by Washington and Stockholm differed in only minor ways),and the EIDCs were unable to reconcile final bulletins throughout most of thetest because of the heavy workload involved. Hence, there were substantialdifferences between final epicenter lists produced by Moscow and the other EIDCs,and fewer (but still significant) differences between those produced byWashington and Stockholm. Perhaps the most notable difference was in the totalnumbers of events located -- Moscow located appoximately 450 whereas Washingtonand Stockholm located somewhat more than 900 events.

S.

While further analyses are needed prior to drawing firm conclusions, thenumbers of located events as a function of magnitude appear to be sufficient to "give a first estimate of the detection threshold of the GSE network, as itoperated during the test. Forming the earthquake recurrence curves in theconventional way (see figure D) it may be seen that for magnitudes in the rangeabove about mb = 5:

log Nc = a + b mb, andlog Ni c + b mb

where Nc is the cumulative number of events larger than a given magnitude and Niis the incremental number of events at a given magnitude. The slope, b, is near V- "-" -1 as usual. The detection threshold is indicated by the magnitude at which the i?

observed numbers of events depart from the formulas above.

As inferred from Figure D, the detection threshold for worldwide eventsduring the GSETT was slightly above mb 4.5,including all effects of seismicstation noise conditions, reporting practices at the stations, data transmissionlosses, and analysis programs and procedures in use at the Washington andStockholm EIDCs. Using programs and procedures in effect at the Moscow EIDC, thethreshold was slightly above mb 4.8.

":: :"" C. Romney

164.4%

..

". :';*4.."-

.= • • . °• .•. -• ." '- ,. ' , ". •'*" •••' o.* ' . ., ,'." • .= .= ." • .. ' ° ." ""16 - ° °

£WASHINGTON

100 £

+ +0. + +

0 0 +

2i A

o 4. 4. AW

AA

U. +.

(6 +

10 A

MAG. S

A4 IL4. Aosco

Qi A0A

10 A+

4.4 4. + +

2 6

MAG. S

INRMNA AN UUAIERCREC UVSFRGET17A

3. SUN SYSTEM DEVELOPMENT

3.1 New data Retrieval Software Available on the SUN Nicrosystem

In an attempt to take advantage of the latest developments in UNIXnetworking, we have developed the software necessary to allow, from our SUNcomputer, access to resources normally available only from a VAX computer. Onelimitation of the moderately priced SUN systems is the lack of high densitydrives which allow reading of Center archive tapes. Libraries exist that permita program on the SUN to use tape drives on another machine on the same local areanetwork. However, before tapes can be read, data requests must be built and todo so requires access to databases. These databases are normally available tousers while logged into the computer that is host to the database. Through theuse of the rsh (remote shell) command we can access these databases and retrievedata across-the local area network.

One script and one program have been developed to handle these tasks. The

script, described in simplified form below, builds a dearchive request from a.. *.given database (RSTN84) for a given event (ekaz.origin) both specified by the

user. In the following example we will assume the user is attempting to build alocal database ekaz.

(get the date for the event from the first column of the origin file)

- set DATE = Nawk"§print $it "ekaz.origin"

(retrieve wftape records from RSTN84 on hugo into file ekaz.wftape)rsh hugo cpout RSTN84 wftape "w.date=$DATE" >ekaz.wftape

(predict arrivals and make waveform request files)predar -aw -s ekaz.segrules - ekaz -t ekaz

(compress waveform request)cmpreq ekaz ekaz

(build dearchive request).dreq ekaz ekaz

When the above script is completed the most important file created will beekaz.wdareq. This file drives the program which dearchives the data. This newprogram, rdetwo (for Remote DETWO), is a version of detwo with subroutine callsto the remote tape library replacing the calls to the normal tape library. Theprogram has the same command line syntax as detwo, except that when a remote tapedrive is desired, the device name must be prceeded by a machine name. Forexample: rdetwo -i hugo:/dev/rmtO -r ekaz. The above command line when issued onthe SUN will-TuiTa a waveTo-rm-Tab-aaasT-ekaz from data read from drive zero onHUGO.

I.-

,.. :%

18

"Il...p ".-. . . . . .. . . . . . . . . . . . . . . . . . . . . . -: .

Since the VAX and the SUN have different representations for 32 bit integers itis necessary to convert the data words before they are written onto disc. Rdetwo irhandles this automaticaly and silently. The end result of this session is asuite of files that comprise a local official Center database. The suite offiles includes a number of ".w" files, that is, files with names like ekaz.l.w,

* "ekaz.2.w, etc. An index for the waveforms exists in ekaz.wfdisc, while thepredicted arrivals are found in ekaz.arrival. Ekaz.assoc links the arrivals toekaz.origin. The arrivals were predicted and waveforms segmented based on thecontent of ekaz.segrules. Ekaz.wftapes contains indexes to all the waveformsarchived for the day the event falls on, while ekaz.wfreq and ekaz.wdareq arewaveform request files, desired and doable respectively. At this point the

" ," database is ready to be used by any number of official Center programs.

M. A. Tiberio

19'

.,-T-'I.-

°°-.-

I: !.: ,...p.d

I. -V

, . ..- .,119 r"-i .'

YIJ -"FI k AATV W WYN; 'R °"°

3.2 Center Standard Software Now Available on the Sun M4icrosystem

With the acquisition of new SUN Microsystems Computers, the processing ofseismic data can now be downloaded from the VAX to the SUN. Since both machinesshare a 32 bit data word and both run Berkeley UNIX version 4.2, transfering .programs is not as difficult as it would be to, say, move the same programs to a16 bit PDP 11. Long integers and floating point numbers have different formatson the VAX and SUN, so any data files in binary need to be converted. Of coursethe two machines have different instruction sets so it is necessary to recompileall programs to their binary executables. With these facts firmly in hand weundertook the task of moving a selected subset of official Center for SeismicStudies software from the VAX to the SUN.

Deemed most necessary were programs that aid the seismologist, not routinedata processers. Since the floating point performance of the SUN2 is poor incomparison to the VAX, no programs that perform large numbers of floating pointoperations, such as detectors, filters and spectral programs, were ported.Certain libraries of seismic routines that support the wanted programs were, ofcourse, also moved. Along with the Center official programs we moved many shell¢ ~scripts that we have written that aid in working with Center official "'-

databases. Below is a description of those programs, scripts and libraries nowavailable on the SUN.

PROGRAMS;fmt.arrival prints format of arrival filefmt.assoc prints format of assoc filefmt.origin prints format of origin filefmt.wfdisc prints format of wfdisc file z ..dareq creates dearchive request from waveform requestpredar predicts arrivalsrdetwo remotely dearchives waveformsrdmparc remotely dumps an archive tape

q trix gives possible origins from three arrivalslocit relocates an event using multi phase algorithmdlaz gives distance and azimuth between two pointsreg gives region number or name

"- '-sta gives station location, etc. --tt gives travel times

SCRIPTS:AAWK reformats arrival fileOAWK reformats origin fileWAWK reformats wfdisc fileORI allows interactive input of origin info.arrival.awk awk template for arrival fileassoc.awk awk template for assoc fileorigin.awk awk template for origin filewfdisc.awk awk template for wfdisc filewftape.awk awk template for wftape filegetr gets dearchive request from databasegetw gets waveforms from archive tapesputw archives waveforms onto tapesmakesegrules makes a segmentation rules file for predartapes tells what tapes are needed for dearchive request

20'-i

LIBRARIES:libextf external files handling routineslibrtap remote tape handling routineslibseis various seismic routines (travel times, dist, az)libtap native tape handling routineslibtime epoch time handling routines

~ .*libwfa wfdisc/wftape manipulation routines

M4. A. Tiberio

.21

4. RESEARCH

4.1 Preliminary Study of mb Bias at Selected Soviet Seismic Stations

Magnitude mb was determined for five earthquakes on 25 and 27 may 1980, fromrecordings at eight Soviet seismic stations on a 4,300 km-long profile fromeastern Kazakh to eastern Siberia. The records were hand-digitized, andmagnitudes were determined from traces corrected for instrument response as wellas the uncorrected traces. Four of the five earthquakes were at Mammoth Lakes,California, in the western Great Basin; the fifth was at Tonga in the southPacific. Magnitude residuals with respect to network-averaged mb listed in theISC Bulletin were postive for stations at Yakutsk and Seymchan in Siberia,negative for raypaths through the Baikal rift zone, and slightly positive for astation at Semipalatinsk, about 100 km from the East Kazakh test site. Acomparison of tabulated magnitude residuals for Soviet seismic stations fromprevious work of Ringdal (1985), North (1976) and Vanek et al. (1978, 1980) showsexcellent agreement between these studies. Our results are slightly morescattered but in good agreement with previous work. Recalculation of mb for 83events recorded on granite at the Nevada Test Site provided a determination ofmagnitude bias( mb= -0.10±0.35) for the NTS granite site site with respect to ISC :...magnitudes. Comparison of this value with mb for the Soviet stations provides adirect measure of the magnitude bias of the NTS area relative to areas in themu ,USSR in which the stations are located. This bias reflects only differences in

attenuation, and does not account for other effects such as differences incoupling, focusing, tectonic release, and instrumental effects. The magnitudebias due to attenuation differences between the NTS granite site and the stationat Semipalatinsk is -0.20. Details are shown in the following table. Thecomplete report is published as Center Technical Report C86-04.

A. Ryall

" " REFERENCES:

North, R.G. (1976). Body ware magnitude mb, Lincoln Laboratory, SemiannalTechnical summary, ESD - TR - 76 - 185, 1-37.

Ringdal, F. (1985). Study of Magnitudes, Seismicity and Earthquake• -. Detectability Using a Global Network, Center for Seismic Studies Rept., in

pressVanek, J. and 32 others (1978). Station corrections for longitudinal waves

in the Uniform Magnitude System of the Eurasian continent, Akad. Nauk SSSR,Izvestiya Earth Phyics, 14, Amer. Geophys. Un., 169-178

C Vanek, J., N.V. Kondorskaya, I.V. Federova and L. Khristoskov (1980).Optimization of amplitude curves for longitudinal seismic waves forpurposes of development of a uniform magnitude system for seismic obervationson the Eurasian continent, Doklady Akad. Nauk SSSR, 250, 834-838, transl.

7 :by Scripta Publ. Co.

22 i

S ~Comparison of Station Residuals and 6Mj(B2-Jh.Nw,)

Sta Ringdal North Van.k' 6m I 46;16m,(0B2-NV)'

BKR +o.38*0.33 +0.30 -0.44

*BOD -0.02*0O.34 .0.05 +0.10 -0.11L

CLL +0.16*0O.26 +0.20*0.32 +0.09 -0.26

ELT +0.15±0.34 .0.01 1+0.15 -0.20

FRU +0.35*0.33 +0.36 -0.46

ILT +0.06*0.32 +0.03 -0.16

IR -0.03±0.31 .0.32 -0.06 -0.04

KHC +0.03*0.26 +0.10±0.26 +0.08 -0.17

jKEE +0.37*0.31 +0.31 .. 4

KRA +0.32*0.29 +0.22*0.29 +0.22 .0.35

KZL -0.06 +0.14 -0.14

MOX +0.07*0.25 +0.02.*0.27 +0.01 -0.13

OBN +0.39.+0.33 +0.39 -0.49

PET +0.24±0.30 +0.35 -0.40 .

PRU +0.04±0.24 -0.06 -0.09 -e

SEI +0.25 +0.28 -0.37

SEM +0.07 +0.10 +0.12 -0.20

TIK +0.03*0.37 +0.00 -0.12 i

UST .0.12 -0.10 +0.01e

VYAK +0.43*0.34 +0.33 +0.42 -0.49 .

YSS +0.20*0.41 +0.02 .0.2 1

ZAK .0.11*0.33 .0.03 -0.03

1I Given by authors as A Mb corrections relative to base station

OBN. Sign reversed and all values increased by 0.39 for comparison with

Ringdal (1985).2.-- Mean of Sm1b for Tonga earthquake plus average . p

.5m for four Mammoth Lake evets from Table 9. Uncorrected data.3 -- Mean of Stab for Tonga earthquake plus average Iim 1 for four Mammoth Lakes events from Table 10. Corrct@A data.

4 - Average residual (.0.10*0.35) for 0B2.NV with respect to ISC

t Bulletin minus average station residual.

23

% %W

* 4.2 Noise Values and Detection Thresholds of GSETT Seismological Stations

Seismic noise levels and detection thresholds for short period teleseismic Pwaves and long period Rayleigh waves have been estimated for seismologicalstations participating in the Technical Test of the Ad Hoc Group of ScientificExperts carried out in the fall of 1984 (GSETT). The estimates are primarilybased on noise and amplitude measurements reported from the stations to theExperimental International Data Center operated at the Center for Seismic Studiesduring the GSETT.

The measurements at the stations of short period noise, (NSZ), consisted ofthe maximum trace amplitude at a frequency between 1.0 and 5.0 Hz or at ai ." frequency close to that of a detected signal together with the associated period .,within 30 seconds before the signal onset. In many instances it was, however,

not possible to measure the noise amplitudes at these specified frequencies.

Analysis of Data From Six U.S. Stations.

Mean values of short period amplitude measurements at participating U.S.Fig.E as amplitude/period ratios and as a function of frequency (i.e., the

reciprocal of period). The plotted values can be thought of as approximate"spectra". An amplitude roll-off inversely proportional to frequency has been

"." drawn for each station for comparison and the "spectral' follow this roll-off over ..some frequency band for all the stations except for LTX and RSSD. The "spectra"

drop from lower frequencies systematically up to some frequency between 1 and 3Hz beyond which a flattening of the amplitude occurs. Statistical testingsuggests that measured noise amplitudes for a given period follow approximatelynormal distributions and the variation of the amplitudes are compatible with astandard deviation equal to the commonly used value of 0.2 regardless of period. 2 -'

Long period noise was usually only measured at the GSETT stations when Wp associated with a detected short period P wave. The long period noise

parameters, (N2LZ), consisted of the largest trace amplitude with a periodbetween 10 to 30s measured on the vertical component within 5 minutes precedingthe initial P wave. Mean amplitude values plotted as a function of frequency areshown in Fig. F for the six U.S. stations. The "spectra" usually have apronounced minimum around the 20s period (or 0.05 Hz).

- The noise amplitudes for the short and long period bands varies withfrequency and in order to determine an operational noise level that representsthe station detection capability it is, therefore, necessary to take thefrequency distribution of signal amplitudes into account. The empiricaldistributions of the period of small short period teleseimic amplitudes are shownin Fig. G at each of the six U.S. stations. For most of the stations the

* - distribution peaks between 1 and 2 Hz. In order to represent each station with. one single noise amplitude the empirical distributions of the wave period as

shown in Fig. G have been combined with the mean noise amplitudes shown in Fig. Eas a function of period. These weighted mean values or "operational" noiseamplitudes are drawn as horizontal lines in Fig. G and are generally lower thanthe mean amplitudes at 1 Hz.

V

24*..

or .FNI."

Plots of the cumulative number of recorded earthquake signals against thelogarithm of the short period amplitude/period ratio are given in Fig.H for theU.S. stations. These plots, which taper off at low values of theamplitude/period ratio, resemble recurrence curves for earthquake magnitudes andcan be used to estimate the amplitude detection thresholds independently of thenoise levels. Both the "operational" noise levels and the estimated amplitudethresholds are drawn in the figure as vertical lines and the gap between the tworepresents the minimum signal-to-noise ratio below which the station will notdetect signals. The following values were obtained for the U.S. stations:

Station Noise Ampi.Thres. 3-W A It- -

code Mean Std Mean Std.. (nm ls - (nm /s-TEX_ Z.Z .ZU A . U., .

LAC 1.6 0.30 11.5 0.34 7.3LTX 1.0 0.18 3.2 0.18 3.2SNY 4.5 0.19 12.0 0.16 2.6

RSON 3.8 0.19 10.0 0.22 2.6RSSD 2.1 0.16 4.7 0.18 2.2

The values of the signal-to-noise ratio, SNR, which in the table are defined asthe ampltiude threshold divided by the "operational" noise level, are all equal '.

* to or larger than the commonly assumed value of 1.5. Similar results wereobtained for several other GSETT stations. This indicates that 1.5 may be low asa standard value for the "average" station even if there may be stations that -

operate at minimum signal-to-noise ratios around this value.

Analysis of Data From Other GSETT Stations ".

The analysis of noise and signal amplitude data illustrated above for the sixparticipating U.S. stations was applied to as many participating GSETT stations

P mas possible in order to compile an updated list of station noise and detectionthresholds for the entire GSETT network. Short and long period noise datareported during the GSETT was used for 48 and 26 of the stations respectively andestimates published in the seismological literature were used for the remainingstations. These updated noise values have been transformed into magnitudethresholds, from a standard magnitude formula assuming a constant Q value (3.73

_ and 3.19 for short and long period respectively) corresponding to average values" of Q in the interval 30 to 85 degrees. The distribution of the thresholds are

shown in the bar chart of Fig.I. The Ms thresholds of long period Rayleigh waveshave been given in equivalent mb values for earthquakes using the relation:mb=0.62 Ms+2.03.

In summary, the following general conclusions can be drawn from the analysis.- of the noise and amplitudes at the GSETT stations.

The revised noise values differ significantly from the preliminary valuesassumed before the GSETT. The revised short period values are either

significantly lower or significantly higher than the preliminary values for mostof the stations. The revised long period values are generally much higher thanthe preliminary values.

25

The GSETT stations constitute a global network with large variation in noiseband detection characteristics among the individual stations. The median value ofthe mb detection threshold as defined above is 4.9 which is a rather highvalue. Most of the stations have a significantly higher detection capability ofteleseismic short period P waves than of long period Rayleigh waves forearthquakes. The variation in short period detection capability among the

stations is well over a magnitude unit whereas the long period detectioncapability has less variation among the stations. The network of GSETT stationsalso has aclearly non-uniform geographical ditiuinwith motof the

stations concentrated in the Northern Hemisphere and in Europe in particular.The stations in the Northern Hemisphere have also generally lower magnitudethresholds than stations in the Southern Hemisphere and more than half of themost sensitive GSETT stations were located in Europe.

* -tA more complete report on the results summarized above is published as CenterTechnical Report C86-03.

H. Israelsson

--

..-

4 . 426_

• ".,'.:.,r.'.,777

AMPLITUDEN(nm/sI

10100 100

FBAS LAC LTX

10 10 10 ,.

1 10 1 10 10 I I I fill-

10 100 100

.1RSNY RSGN IRSSD

10 10 10

l1110 9 ! I i IUt? 1 10 0.

FREQUENCY (Hz)

FIGURE E ESTIMATED MEAN VALUES OF NOISE AMPLITUDES (AMPLITUDE/PERIOD)AS A FUNCTION OF FREQUENCY. THE 95% CONFIDENCE LIMITS ARE INDI-

OBSERVATIONS. THE SLOPING LINES CORRESPOND TO AN AMPLITUDEROLL-OFF PROPORTIONAL TO FREQUENCY SQUARED.

27

. -Mrs

AMPLITUDE5 15

(nm/s)

10FBAL 10LAG LTX

CCD

CDQ

1010 10(OLD

0.10 0.10 0.10

15,1 I 5I11 1 I 111.11 1 1 1 1 I lliil I I Jiml

ISN Y 'RSON PRSSP100 Q 100 100

CD C

10 10 10

0.10 0.10 0.10

FREQUENCY (Hz)

N.5

FIGURE F ESTIMATED MEAN VALUES OF NOISE AMPLITUDES (AMPLITUDEIPERIOD) AS AFUNCTION OF FREQUENCY. THE 95% CONFIDENCE LIMITS ARE INDICATED BYHORIZONTAL BARS IN CASES WITH SUFFICIENT NUMBER OF OBSERVATIONS.

28 i

~.. NO. OF

(m /ns) '''' I I I 1 1111a a 111

10 )FBAS 10LAGC 0 LIX

10 10 10

100, 10 10

X XC

X

1 10 1 10 110

100 ~RQUNY 100HzIN 10

X 1%

FIUR X NUBE OFDTCE1HR0PRO EEESI PWVSWT MLAMLTUE ASAFNTOMFMAUE OIATFEUNY H

MEA NIS LEEL HVEBEE DAW FO CMPRION.AN ETI

.5 X

10lo1I ~~~ ~ ~ ~ ~ * I, I -id I I1111 il:

I~~~~~ ~ ~ ~ ~ I illliij~ I 1111 1 1*T1 11 lah11l

FBAS L 1SNY100 100L

10 10

L, 1?1 1 lo 1il 1 Ilitf I 1till11

10LTX Q6 BSON

10 10

110 i0 0 1 .10 100 Q

LAC 1RSSD100 w 100 OEA

-j .*W NOISE

a~~ ~~ 00 THEHL00

10 1010 0 o

AMPLPUDETUDEs

7r DEECIO THRESHOLDSADOPRTOA'NIELVSASEFEDN

SAFCTION F ANT8AE DAWNLIASDEVERICD ALINESTORACOMPARIN.

30

* U

-01'7.

0~0

-4,, ,.3

4'. ,- ' -.:

No a..

0

-Lo

,: , , ,'.

~0 *

0

%",

4-. I - * 43 0

a,0 0

00 "01 008 009 00 oh 00 9z 0 u ',su Illt)O*O

31

, 'a . - ". " "-.'- . " , " d ' '- .' ' . ". ' ' . ". "i 1 . ':.

II.

.,,,-8

4k

...I -rt t'+,.'u ~ '' L ' c 'P ' +",'','+_-;,,t. + '. "" + "" ""• "" "" " "" """" "" " "+ ""